deptula: these individual little small sensor balls could generally see only a very small area on the earth even if you're up at 20,000 feet. chad: i can't look out and see that maybe there's something else going on outside and around me. narrator: so the crew's grasp of the situation can be limited and when they zoom in, they sometimes lose sight of key details on the ground. this secret 2010 transcript obtained by journalists through the freedom of information act reveals what can go wrong. while supporting forces on the ground, a predator team mistook a convoy of afghan civilians including women and children for militants. the pilot asks "is that a... rifle?" the answer is, "can't really tell." after hours of monitoring, the ground commander calls in a strike...

and 23 civilians die in part because of limitations of the sensors. but engineers are working to create new sensors that can see more, in greater detail, than ever before. deptula: today we've developed sensors that can watch with an "all-seeing" eye and see an area about the size of a small city all at one time. yiannis antoniades: this is the next generation of surveillance. for the first time we actually have permission from the government to show the basic capabilities. it is important for the public to know that some of these capabilities exist. narrator: engineer yiannis antoniades designed the new sensor, known as argus. with 1.8 billion pixels, it's the world's highest resolution camera. argus fits inside this pod that attaches to the belly of a uav.

but because much of the work is classified, we can't see the sensor itself. antoniades: because we are not allowed to expose some of the pieces that make up the sensors so you get to look at pretty plastic curtains. narrator: also known as "wide area persistent stare," argus is the equivalent of having up to 100 predators look at an area the size of a medium-sized city at once. this image was taken 17,500 feet above quantico, virginia and covers 15 square miles. this whole image is at a very, very fine resolution so if we wanted to know what is going on in any spot along this image, say near this building, at this intersection we can generate a moving image that shows what's going on in the area. narrator: simply by touching the screen, antoniades has opened up a window showing a detailed area

while still maintaining the broader context. antoniades: and everything that is a moving object is being automatically tracked. the colored boxes represent that the computer has recognized the moving objects. you can see individuals crossing the street, you can see individuals walking in parking lots. there's actually enough resolution to be able to see the people waving their arms or walking around, what kind of clothes they wear. and you could pick the location of where you produce these images anywhere in the entire field of view. narrator: antoniades can open up to 65 windows at once and can see objects as small as six inches on the ground. antoniades: from even 17,500 feet, the white thing that you see flying around is a bird. narrator: argus streams live to the ground and also stores everything-- a million terabytes of video a day, which is the equivalent of 5,000 hours of high-definition footage.

antoniades: so you can go back and say "i would like to see what happened "at this particular location three days, two hours, four minutes ago," and it would actually show you exactly what happened as if you were watching it live. narrator: to create the world's highest definition camera, antoniades needed to design a new imaging chip but darpa, the project's funder, wanted to move fast and keep costs down, so he borrowed technology that most people have in their pockets. antoniades: inside this cell phone we find a tiny little camera. so if you were to take off the majority of it you'd be left with an imaging chip. if you were to take 368 of these and make a big mosaic out of them and start shooting images now you have argus. narrator: unlike the predator camera that limits field of view, argus melds together video from each of its 368 chips

to create a 1.8 billion pixel video stream. this makes it possible to zoom in and still see tremendous detail. whether argus has been deployed in the field is classified. antoniades: i'm not at liberty to discuss plans with the government. but if we had our choice we would like argus to be over the same area 24 hours a day seven days a week. that's not very easily achievable with manned platforms. this is where uavs come in and they're absolutely the perfect platform. narrator: argus may be mounted on an armed uav like the predator, a long-range platform like the giant global hawk or a development craft called the solar eagle that may someday stay aloft for years at a time. cummings: the u.s. air force right now has the ability to archive every single video that comes off

of every single uav. we're moving to an increasingly electronic society where our movements are going to be tracked. narrator: while technologies like argus are expanding the reconnaissance power of uavs, unmanned aircraft remain vulnerable in other ways. the air force's large drones still crash more often than manned planes. the $100 million global hawk is nearly three times more likely to crash than the u-2 it was designed to replace. even if there's no pilot to lose, when a drone goes down it can still be a problem. deptula: it doesn't take a thermonuclear brain surgeon to figure out that if you are operating a highly advanced, technologically capable system that you'd prefer not for it to fall into adversary hands. narrator: in 2011, an american spy drone the rq-170 sentinel,

was lost over iran. there was a problem with the aircraft, and it landed in an area it shouldn't have landed in and that's about all i'm going to say. narrator: the iranians claimed they hacked into the drone's control system and took over the craft. and in theory it is possible to take over any computerized system. singer: you could never call up maverick, in his f-14, in the movie top gun, and say, "maverick..." talk to me, goose. singer: "recode all american jets as russian jets." tom cruise would have laughed his weird cackle and said, "no way, man." with a computer, you can do that-- either jam it, or even more so co-opt its operations. and it's a whole new era in war. it opens up a lot of new possibilities and, of course, huge new dangers to think about. narrator: why the sentinel went down

remains classified. but experts point out that drones often crash for very basic reasons. uavs sometimes aren't that smart, their self-diagnosis isn't all that smart and by the time they have a problem, it's too late. you lose a link, lose power, and you're gone. narrator: control can be lost for a number of reasons. when reaper pilot chad simply banks too sharply he loses the satellite link. chad: uh-oh. narrator: the picture freezes, and he's momentarily flying blind. chad: that was me turning. (system beeps) narrator: he quickly levels the craft and restores the link. chad: i was turning aggressively and it had trouble keeping up the satellite link. deptula: there is a degree of vulnerability involved with remotely piloted aircraft that have a command link-- where they're actually operated by operators-- that you can overcome by having a human in the cockpit. pilot: dark star, dark star...

mcdonough: if for some reason i all of a sudden hear an increase in airspeed, i start to feel a roll and i'm not intending for that to happen well, something is going wrong and i need to make sure that i am doing what i need to be doing. narrator: unlike manned planes, drones depend on control links that can be lost or potentially as may have happened with the sentinel, even taken over by the enemy. but what if a craft could operate on its own free of any links, and even make its own decisions? in a lab at the university of pennsylvania, vijay kumar is funded in part by the military to create autonomous drones that don't need external links and, like us, can sense their environment. what you see on this robot are these two chips here which are essentially rate gyroscopes. these play the same role as the semicircular canals in the human body located near the ears, which essentially tell us orientation. so the rate gyroscopes that are on board can actually measure

these angular velocities at thousands of times a second. this chip here is the accelerometer and this allows the robot to sense accelerations in the lateral direction so these are analogs to the otolith organs that measure acceleration in the human head. narrator: when a human pilot feels an unexpected change in acceleration, he knows to adjust the aircraft. kumar: the robots do exactly the same thing. narrator: the sensors adjust the craft by changing the relative velocities of the rotors and allow the drones to follow a leader with precision. kumar: a fundamental problem in coordinating multiple robots is the ability to maintain formations. what a robot has to do is determine where its neighbors are and figure out what the relative position is and then monitor the relative separation very carefully. you only need to tell one robot how to move and the other robots essentially

maintain formation by just keeping a specified relative distance. in the figure 8, they come within inches of each other so they have to combat aerodynamic effects from their neighbors and they have to have very, very precise control. and all that is done autonomously. narrator: the precision of the robots allows them to do some things more quickly and accurately than human pilots can, like predict the movement and the shape of an object and adjust accordingly. kumar: in terms of acrobatics i think it'll be hard to beat what a robot can do. the neuromuscular system in the human body, there may be delays of the order of 80 milliseconds or 200 milliseconds before you actually take an action in response to what you see. well, robots have this unfair advantage. they can do these computations hundreds of times a second. so your delays are of the order of a millisecond, and perhaps even less. narrator: in the lab, the drones

communicate with a central computer that uses motion capture-- an optical system that tracks silver reflective markers on the robots and tells them where they are at all times. but soon it may be possible to cut the cord. and they've already developed another autonomous drone that can go anywhere on its own. the holy grail is to do all of this without any kind of external sensing, without gps and in principle, we can do it. these bigger robots actually rely on observations of external features to tell them where they are in the environment. narrator: this drone carries a laser range finder that determines distance to obstacles and a depth camera that reveals 3-d information about the surroundings. kumar: they carry on board the processing power, the sensors that are necessary to look at the environment to reason about the environment. so they can take their relative location

and the location of the features to build a three-dimensional map. i'd like to see this technology being used for humanitarian purposes. imagine there is a 911 call from the building. i think we will soon have the technology that enables, let's say, 20 uavs to just swoop through the building and within a minute find out who is in each room and then communicate that to firefighters who are waiting outside. but any technology that you develop, there are always people who are going to use it in ways that the designer never intended them to be used. narrator: and the key components of drone technology are already available to virtually anyone. man: go over there. man: i got into the hobby about four years ago. it started out as a toy for my dogs to chase. i'm pretty obsessed with the hobby. i am currently single, have been for 30 years.

no kids, i just live here with my three dogs and i get to spend nearly every night and every weekend working on these things. narrator: corey brixen of orange county, california is one of tens of thousands of drone hobbyists across the country. brixen: the basic quadcopter is about $500, which is the frame the motors the speed controllers, the electronics that you need. this is actually the gps sensor and that's sending the signals out to the satellites. the cameras, as small as they are they're high-definition footage. narrator: like the predator, corey's quadrotor has a programmable autopilot that uses gps to navigate. he can set locations altitude and speed so the craft can fly autonomously. corey flies for fun, but drones small and large are increasingly used domestically for more serious purposes. the border patrol searches for illegal migrants with predators. police departments are turning to smaller drones

for reconnaissance and thousands of individuals use them for their own sometimes political, ends. we don't consider ourselves to be animal rights activists. we're investigators. some of our targets are pigeon shooters in pennsylvania. so we had the copter up filming what's going on. there were a couple of shots and it's slowly losing altitude, and then it takes one more shot and it's down in the trees. you can hear the shooters start cheering and laughing and all that kind of thing. what they didn't know is that we had more. (laughs) narrator: this incident is the first known shoot-down of a domestic drone. but it may not be the last. the spread of drones has raised concerns about privacy

and even led one senator to propose legislation limiting domestic use. rand paul: i think anybody that can use a device to peer into your activities even if it's from the air outside your window or the air above your property i think there is a right to privacy. and i'm all for making sure that, you know, all these drones don't come back from afghanistan and iraq and get put to a purpose here that isn't consistent with our constitutional bearings. the worry is that in two to three years we'll have 30,000 drones crisscrossing the sky, accumulating all this information. it's a game changer, but it's not just an american revolution. there are more than 55 other countries right now that are building, buying or using military robotics particularly in the air. narrator: iran recently announced it has its own armed drone with a range of over a thousand miles. it's not known whether any of the technology was adopted from the lost u.s. sentinel.

but in the drones arms race, the u.s. is constantly developing new technologies, like a small flying camera with a twist called switchblade. the switchblade will launch out of this tube and the wings, the tail, the propeller will all spring open; that's why it's called switchblade. and so when the operator finds the target by looking into the viewing screen, he will then send a command to the switchblade that will enable the switchblade to basically hone in on that target and detonate its explosive charge upon contact. it's a tool that our customers are very excited about and it's a tool we think will protect our forces and help save innocent lives. narrator: switchblade's already been deployed. and a killer drone that may someday replace a manned fighter is on the horizon. it's really about a decade ago some folks got together and they said, "what if we want to put "unmanned aircraft on an aircraft carrier?

how do we do that?" narrator: this x-47b prototype is the world's first tailless fighter-sized drone. but it's still a year or two away from its first goal: to take off and land on an aircraft carrier. matt funk: we're maturing the technology required to take this unmanned vehicle to land on the flight deck autonomously. we're focusing on the mechanics of landing the aircraft where you have a predetermined position that you're targeting to and the aircraft can react very quickly to changes in that environment to put itself on that preplanned position. the system has a lot of sensors, a lot of instrumentation on it so it knows how all of its subsystems are behaving, it knows how its engine and all of its control surfaces are performing at any time. it'll turn where we expect it to turn it'll come back when we expect it to come back and it'll land on the runway that we planned. narrator: drones are no longer just eyes in the sky and someday uavs like the x-47b may be involved in sophisticated attacks alongside manned jets.

it's the idea of the mix the team working together. that's probably the future. we won't see every plane on the aircraft carrier be one of these. so right now, the vision is, oh, we'll just have one or two but when they get them and maybe they prove more useful then it may be: "well, we don't want one or two, we want a squadron." it's a lot like if you look at the first use of mechanized forces. originally it was lots of horses and just couple of trucks and tanks, and you saw things change and change and change over time. we'll probably see the same thing happen with robotics. narrator: as drones like the x-47b mature, they'll become more autonomous likely conducting bombing runs then more complicated air-to-air operations. even as policymakers debate the rules of engagement for drones technology is moving forward toward a time when drones might operate with intelligence that more closely resembles our own. but for now, they can't do what a pilot can.

funk: when you get an aircraft like this over hostile territory, you'll have potential targets or threats that will pop up unplanned. so what we haven't done yet is develop the technology that would then react to those unplanned targets or threats. thank you very much for coming to pax river today. eremenko: i think uavs of the future will certainly be able to exhibit increased levels of autonomy. but i think if you were to ask most autonomy researchers, or most ai researchers about whether the rise of the machines type scenario is a real concern, their response would be, "we should be so lucky." in fact, if we could get little slivers of that kind of adaptive and cognitive capability into systems, that would be a very significant breakthrough over where we stand today. mcdonough: i've flown about 2,000 hours and the missions that i've flown seldom go as planned. there's a lot of pieces that you cannot plan for. karem: nothing can replace the human

being in a tactical environment saying what to do next and why not this and why not that. karem: an airbus 320 with 150 passengers goes into a flock of canadian geese and needing to land... karem: "i can't do it-- i'm going on the hudson." you want to be in a robotic 320 in that situation, or do you want sullenberger to land you? because you can't anticipate everything. narrator: this ability to respond to the unknown may be the final hurdle if drones are ever to fully replace manned planes and start making decisions on their own. karem: i think we're far, but let me say i am the last guy who says impossible. narrator: as human ambition drives innovation forward the only thing that's certain is that the predator and other drones

of today are nothing compared with what's to come. ultimately darpa's mission is the creation and prevention of strategic surprise, so if we are successful, we will all be surprised. chad: historically you look at when the wright brothers first flew in 1903. a hundred years later, we are actively flying remotely piloted aircraft. so we're kind of on the ground floor now. there's nowhere to go but up. they called him "lucky lindy." then the ultimate tragedy struck. man: these are hardcore guys, daring enough to kidnap the lindbergh baby and risk the death penalty. but was justice served? two people, three people... for sure, it is not one person. and what can modern forensics reveal?

man: a computer can do a lot more than a document examiner can do. "who killed lindbergh's baby?" next time on nova. major funding for nova is provided by: supporting nova and promoting public understanding of science. and by the corporation for public broadcasting. and by contributions to your pbs station from: additional funding from: inspiring tomorrow's engineers and technologists. and millicent bell through: on pbs its a stunning new season starting with the impossible to resist downton abbey. violet: nothing succeeds like excess. and an engrossing story of defiance and courage from american experience. douglass: it is not light that is needed, but fire! and a series filled with role models for all women clinton: we were breaking new ground.

this season, there's only one place that can take you anyplace. attenborough: isn't it wonderful? pbs the exploration continues on nova's website, where you can watch this and other nova programs see expert interviews, interactives video extras and more. follow nova on facebook and twitter and find us online at pbs.org/nova. this nova program is available on dvd. to order, visit shoppbs.org, or call 1-800-play-pbs. nova is also available for download on itunes. captioned by media access group at wgbh access.wgbh.org

froggatt: the dining room scenes certainly are very intricate. bonneville: as any dp director and production manager will tell you, they are complicated. naeme: they're very important scenes. they're pivotal in the drama because it is the main occasion when people in a family or any kind of society can come together and discuss things while they're eating a meal. mcgovern: it's definitely one of the more grueling parts of shooting "downton abbey", because you have to cover every single character at that table, and so you have to do it over and over and over again. there are no shortcuts. it's all to do with eyelines. you can't just sort of put one camera on and two cameras and hope for the best. and you have two lines looking to the person to your left

two lines to the person opposite you. everyone is looking at different people around the table. and every single one of these shots every single one of these angles has to be covered so they're very time-consuming. froggatt: all the food has to be brought out, and it has to look hot, you know. the continuity of that is a mine field. in season 1, we, i think learned our lesson because i think we had some quail or something like that which, starting shooting at 8:00 by 3 p.m, that quail's fairly manky. i come in through the door into that dining room, and with all the lights going and all that fish on the table and 20 actors and 30 crew all in the room, it can be a slightly rank smell. [chuckling] fish, for example, is slightly off the menu. ha ha!

er